Evolving dispersal and age at death

Calvin Dytham and Justin M. J. Travis

Dytham, C. and Travis, J. M. J. 2006. Evolving dispersal and age at death. � Oikos 113:530�538.

Traditional, and often competing, theories on ageing agree that a programmed age atdeath must have arisen as a side effect of natural selection, and that it can have noadaptive value of its own. However, theoretical models suggest that ageing andprogrammed death can be adaptive. Travis J. M. J. suggested that if fecundity declineswith age, a programmed age of death evolves through kin selection and that the natureof dispersal is crucial as it determines the degree of spatial structure and hence thestrength of kin selection. Here, using a similar model, we consider the interplaybetween dispersal and age of death. We incorporate more realistic dispersal kernels andallow both dispersal and age of death to evolve. Our results show each trait can evolvein response to the other: earlier age of death evolves when individuals disperse less andgreater dispersal distances evolve when individuals are programmed to die later. Whenwe allow dispersal and age of death to evolve at the same time we typically find thatdispersal evolves more rapidly, and that ageing then evolves in response to the newdispersal regime. The cost of dispersal is crucial in determining the evolution of bothtraits. We argue both that ageing is an overlooked ecological process, and that the fieldof gerontology could learn a lot from evolutionary ecology. We suggest that it is time todevelop the field of ecological gerontology and we highlight a few areas where futurework might be particularly rewarding.

C. Dytham, Dept of Biology, Univ. of York, York, UK, YO10 5YW (cd9@york.ac.uk).� J. M. J. Travis, Centre for Ecology and Hydrology, Hill of Brathens, Banchory,Aberdeenshire, UK, AB31 4BW.

There are three generally accepted theories of ageing:

mutation accumulation (Medawar 1952), antagonistic

pleiotropy (Williams 1957) and disposable soma

(Kirkwood 1977). All three of these theories in some

way imply that ageing is the consequence of evolution

acting to promote fitness early in life at the expense of

fitness later in life. Mitteldorf (2004) summarised these

traditional views on ageing writing: ‘‘Standard theo-

ries . . . regard senescence as an epiphenomenon of

selection, with no adaptive value of its own.’’ Recent

discoveries in nematodes (Kenyon et al. 1993, Kimura

et al. 1997, Friedman and Johnson 1998), insects (Bartke

2001, Tatar et al. 2001) and mammals (Migliaccio et al.

1999, Holzenberger et al. 2003, Lithgow and Gill 2003)

of genes that, when mutated, increase life span challenge

these conventional theories, and there is a need to

reassess the merit of Alfred Russel Wallace’s early ideas

regarding adaptive death (Wallace 1889). Travis (2004)

demonstrated that in a spatial model a gene that results

in deterministic death at a specified age can evolve, and

does so through kin selection. The spatial scale of

dispersal is a key determinant of the age of death that

evolves (Travis 2004), and in this paper we consider in

more detail how dispersal influences the age of death to

evolve. Specifically we are interested in the evolutionary

interplay between dispersal and age of death. We are

aware that, despite some recent interest in the ecological

consequences of ageing (Bonsall and Mangel 2004,

Hendry et al. 2004, Broussard et al. 2005), for many

ecologists theories on ageing may be a new topic. First

Accepted 28 October 2005Subject Editor: Tim Benton

Copyright # OIKOS 2006ISSN 0030-1299

OIKOS 113: 530�538, 2006

530 OIKOS 113:3 (2006)

we outline the traditional theories, and then give more

detail on novel perspectives on adaptive death.

Why do organisms age and die?

This question has long vexed biologists. Alfred Russel

Wallace first suggested that ageing and death might be

adaptive (Weismann 1882, Wallace 1889). In the 1860s

Wallace wrote ‘‘Natural selection . . . in many cases

favours such races as die almost immediately after they

have left successors’’. Despite some early support, this

adaptive view of ageing and death was soon dismissed,

to such an extent that in the 1920s it was labelled a

‘‘perverse extension of the theory of natural selection’’

(Pearl 1922). This has remained the case since with

almost all biological gerontologists believing that ‘‘long-

evity determination is under genetic control only indir-

ectly’’, and that ‘‘ . . . ageing is a product of evolutionary

neglect, not evolutionary intent’’ (Olshansky et al. 2002).

Today, there are three largely competing theories used to

explain ageing; mutation accumulation, antagonistic

pleiotropy and disposable soma. The evolution of ageing

has been reviewed by Gavrilov and Gavrilova (2002),

Goldsmith (2003) and Kirkwood (2005). Below, we

summarise the key theories on ageing, and for compar-

ison also describe the main drivers of dispersal.

Traditional theories

The mutation accumulation theory for ageing (Medawar

1952) suggests that ageing occurs due to the accumula-

tion of deleterious mutation that result in negative effects

late in life. Later age classes contribute less to lifetime

fitness and therefore mutations acting late in life

are under weaker selection. Antagonistic pleiotropy

(Williams 1957) is an alternative theory suggesting that

ageing is the result of genes that have beneficial effects

when an organism is young, but also have adverse effects

when the organism is older. This theory relies on the idea

that the beneficial fitness effects early in life will be under

stronger selection than the adverse effects later in life.

For some species, it is likely that genes with compara-

tively small early-life fitness benefits can be selected even

if they cause a major reduction in late-life survival.

A third theory of ageing is disposable soma theory

(Kirkwood 1977) which posits that organisms have to

budget their energy among metabolism, growth, activity

and reproduction. Repairing damaged proteins and

error-checking in DNA replication both make demands

on energy resources, and allocation to these resources

has to be balanced with that to other functions. Failure

to do a perfect job in these and other repair and

maintenance functions leads to an accumulation of

damage, which results in ageing.

Adaptive death

The widely accepted view that ageing has no adaptive

value of its own has been challenged (Guarente and

Kenyon 2000, Skulachev 2001, 2002, Goldsmith 2003,

Mitteldorf 2004, Travis 2004). Two individual-based

models provide theoretical support for the concept of

adaptive ageing and programmed death (Travis 2004,

Mitteldorf 2006), the latter invoking a form of group

selection and the former kin selection.

Mitteldorf (2006) developed a lattice model in which

an array of sites is connected by occasional dispersal.

Each site can support a population whose dynamics are

simulated using the discrete logistic equation. Simulation

experiments were conducted in which individals carry

genes that determine their rate of ageing. Mitteldorf’s

results indicate that parameter space exists where ageing

is selected. He suggests that ageing is selected at the

population level as an adaptation that stabilises popula-

tion dynamics. Within a population, selection favours

individuals leaving the most offspring (i.e. have a lower

rate of ageing). However, this individual-level selection

will lead to an increase in the population’s growth rate

and potentially drive the population into chaotic fluc-

tuations which increases the risk of local extinction. On

average, sites where individuals age more slowly will go

extinct more often, and those vacant sites will, on

average, be colonised by individuals with more rapid

ageing. There is a tension between selection at the level

of the individual and at the level of the population. In

many respects, Mitteldorf’s argument represents classic

group selection but with the twist that selection acts at

the higher level to promote demographic homeostasis.

Travis (2004) developed a rather different spatial

population model to explore the evolutionary ecology

of programmed organism death. As in Mitteldorf (2006)

a lattice is used to represent space, but in Travis (2004)

each site on the lattice supports a single individual

(rather than a population). The model simulates a

viscous population rather than a collection of subpopu-

lations each of which is freely mixing. Results from this

model demonstrate that if an individual’s reproductive

fitness declines with age, then an age of death gene is

selected through kin selection. By dying, an individual

frees resources (in this case, space) that can be used by a

younger, fitter individual. If dispersal is local, such that

local spatial structure develops, the individual that

benefits from another’s death is likely to be a relative.

Therefore if dispersal is local, a critical age exists at

which an individual maximises its inclusive fitness by

dieing. The age of death that evolves is sensitive to

dispersal distance, although even for relatively long

distance dispersal, individuals are selected to have a

finite lifespan (Travis 2004). The interplay between

dispersal and age of death is likely to be relatively

complex, and will be mediated through the spatial

population dynamics. Travis (2004) only touched on

OIKOS 113:3 (2006) 531

this issue and in this paper we aim to extend this first age

of death model to incorporate the evolution of dispersal.

Drivers of dispersal

A number of mechanisms have been suggested to play an

important role in driving the evolution of dispersal

strategies. Here, we outline these mechanisms, and

consider how they may together shape the dispersal

behaviour of individuals within a population.

Increased dispersal is likely to be selected for if local

population extinctions are likely. A population will only

survive in the long-term if it is able to recolonise patches

from where it has become locally extinct. The probability

of local extinction will depend upon a combination of

temporal variability in habitat quality, demographic

stochasticity and the nature of the population dynamics.

Extreme temporal variability in habitat quality may

deterministically drive local populations extinct, and in

this case a population will only survive if it is able to

track the dynamic habitat matrix. There is good

empirical evidence that the temporal persistence of

habitat patches is a determinant of dispersal behaviour

for wing-dimorphic insects (Roff 1986, 1994). Local

populations may go extinct through demographic

stochasticity even when habitat quality remains constant

through time. For populations prone to frequent local

extinctions through demographic stochasticity we would

expect higher rates of dispersal to evolve. Demographic

stochasticity will be most important when effective local

population sizes are small. Population dynamics can

play an important role in determining a population’s

persistence within a patch. For example, populations

that exhibit intrinsically unstable dynamics may become

locally extinct very frequently, driving selection for high

rates of dispersal.

At least two mechanisms can promote dispersal even

when local populations are not at risk of extinction.

Hamilton and May (1977) demonstrated that even in

spatially and temporally stable habitats dispersal is

selected. They demonstrated that dispersal is favoured

because it reduces competition between kin. Dispersal

may also be selected for as a mechanism reducing the

effects of inbreeding depression (Bengtsson 1978, Perrin

and Mazalov 1999, Cockburn et al. 2003). Whatever the

mechanism, it is expected that higher dispersal rates

should evolve when local population equilibrium den-

sities are low, for example when habitat patch sizes are

small, because kin competition is more intense and

inbreeding more severe in small populations.

Local population extinction, kin competition and

inbreeding all act to select for increased dispersal. The

evolution of increased dispersal is always constrained by

one or more costs, for example mortality risks during

travelling, investments in dispersal morphology, or a

tradeoff between dispersal and competitive abilities.

Linking age of death and dispersal in a simple patch

occupancy model

Here, we construct a simple patch occupancy model, in

which fertility declines with age and an individual’s

genotype determines both its dispersal propensity and its

programmed age of death. We look first at how different

dispersal strategies influence the age of death that

evolves, second at how different age of deaths influence

the evolution of dispersal, and third at how the two

evolve.

Method

We develop an individual-based, simulation model to

investigate the evolution of programmed age at death

and dispersal distance. Simulations take place in an

arena of cells with individuals located in continuous

space. For simplicity, each cell can only support one

individual. The model is event-based with events taking

place in series. An individual and a type of event are

selected at random and time is incremented by an

infinitesimal amount in a manner similar to that

described in Renshaw (1991). Events are either deaths

or birth and dispersal. This approach overcomes poten-

tial problems with the ordering of events (discussed in

Palmer 1992). We consider a species that has only one

dispersal event in its life and this occurs immediately

after birth. The model is agent-based with each

individual carrying an exact location in space (x, y co-

ordinates), its age, a dispersal distance assigned to any

offspring and an age at death. Individuals can suffer

mortality form a) a background mortality rate indepen-

dent of age or density which is usually set at 0.01 (i.e. 1%

probability of death in a ‘year’); or b) through senescence

when an individual exceeds its assigned age at death.

Following Travis (2004), births occur with a probability

that declines with age and dispersal immediately follows

birth. Organisms are iteroparous and in each birth event

there is only one offspring ‘seed’. The seed moves in a

random direction from the adult selected from a uniform

random variate between 0 and 2p radians. The distance

is drawn from a negative exponential distribution with a

mean taken from the genotype of the individual. So, an

individual that has a birth event may have a dispersal

distance of 4.2 which means that any offspring will move

in a random direction a random distance drawn from a

negative exponential distribution with a mean of 4.2.

Any individual that fails to leave the parental cell is

moved exactly one unit in a cardinal direction from its

532 OIKOS 113:3 (2006)

parent’s location (Fig. 1). A disperser arriving at an

occupied cell is not able to establish.

As this is an event-based model there will be some

individuals who have several events occur to them in a

‘year’ while others have no events. Events occur at

random, so they will be Poisson distributed amongst

individuals.

The arena is a torus so any individual leaving the

arena will reappear on the opposite side. We also use a

mortality associated with dispersal. Following Murrell

et al. (2002) this is applied as a linear, increasing

mortality with distance during dispersal. A value of 20

equates to a mortality of 100% when moving a distance

of 20 units, or 50% for 10 units, and so on. In the

simulations presented here we have dispersal mortality

values of 5, 10 or � (no mortality during dispersal).

There may be mutation associated with a birth event.

Mutations are rare (usually 0.3%). Mutations can affect

either age at death or dispersal distance. Mutations to

age at death are drawn from a uniform distribution from

�/3 to �/3. Mutation to dispersal distance is also

uniform from �/0.5 to �/0.5. Both characters have a

limit at zero and any mutation that gives a value less than

zero is set to 0.01. An individual may only have a

mutation to one character. While there is no explicit

calculation of the relatedness of individuals limited

dispersal leads to population viscosity and the clustering

of individuals sharing a genotype (i.e. kin).

Simulations were carried out on square grids of 200

cell side lengths (40 000 cells) with approximately 1/3 of

cells occupied at the start. Individuals are assigned either

a) random age at death (between 1 and 100), dispersal

distance (between 0.01 and 10) and current age (between

0 and 100) at initialisation or b) set values for one or

both characters.

Simulations are run for up to 150 000 years and where

confidence intervals are shown there were ten realisa-

tions of the model for each parameter value.

Results

When dispersal distance is fixed and age at death

allowed to evolve there is a clear positive relationship

between evolved age at death and dispersal distance

(Fig. 2a). Although the mean evolved values for age at

death for the population are very consistent between

realisations there is a range of evolved values remaining

within a single realisation even after a long simulation

(Fig. 2b). Clusters of relatives (i.e. individuals sharing a

common strategy) are clearly seen as patches sharing a

common age at death.

When age at death is fixed and dispersal distance

allowed to evolve there is an increase in dispersal

distance with age at death (Fig. 3) although the relation-

ship is not linear.

Fixing both characters for all individuals in the

population allows the effect of age at death and dispersal

distance on total population size to be seen. Within a

single dispersal distance, as shown in Fig. 4, there is a

much lower population size with early age at death, this

rises rapidly to a peak and then declines slowly as age at

death increases further.

When both age at death and dispersal distance are

allowed to evolve with no cost to dispersal the dispersal

distance rises quickly and the age at death slowly with

no apparent limit. Simulations run for a very long time

(�/250 000 years) still show age at death drifting slowly

upwards. Mortality during dispersal keeps dispersal

distances down and higher costs to dispersal reduce

both age at death and dispersal distance.

If all individuals start with identical strategies and

both characters are allowed to evolve the population

means reach the same values regardless of starting

values. This is shown for 11 trajectories from different

starting values in Fig. 5a and, in more detail around the

equilibrium value in 5b. Using this range of starting

conditions we only found a single attractor. It is clear

from the trajectories that dispersal distance is under

stronger selection pressure than age at death except when

age at death is very low. We determined this as a

proportion of the final change in a character that

was achieved in the first 1000 years. This is because

old individuals are under very limited selection pressure.

x

(a) (b)

(c)

Fig. 1. This shows a stylised version of the event-based model.In (a) a dispersal event (arrow) fails because the destination cellis occupied. In (b) an individual dies. In (c) a short distancedispersal event (dotted arrow) fails to leave the cell, so adispersal distance of 1 (arrow) is selected for a successfuldispersal event.

OIKOS 113:3 (2006) 533

Figure 5c and 5d show that at equilibrium the

mutation�selection balance is allowing quite a large

variation between individuals.

Sensitivity analysis

Results are not affected by mutation rate or mutation

ranges except in the speed at which an equilibrium level

is reached. Changing boundary conditions to absorbing

or reflecting boundaries also has little effect, except in

arenas much smaller than those reported here and when

mortality during dispersal is very low and more indivi-

duals are lost from the edge of the arena.

Adjusting the background rate of mortality and

changing the shape of the decline in reproductive fitness

with age both have the same qualitative impacts as

reported by Travis (2004). When reproductive senescence

is more rapid, selection favours an earlier age of death.

As background mortality is increased, selection acts to

increase the programmed age of death. This can be

explained by considering the population dynamic con-

sequences of higher background mortality. As extrinsic

mortality is increased population abundance declines

and this reduces the competition of space. This in turn

60

50

Mea

n ev

olve

d ag

e at

dea

th

40

30

20.5 1.0 1.5 2.0 2.5 3.0 5.0 6.0 8.0

Distance7.04.0

(b)

(a)

Fig. 2. The evolved age at death after 10 000 generations. (a)shows mean and 95% confidence interval of 10 realisations ofthe model for a range of fixed values for dispersal distance.Dispersal mortality is 10 throughout. The dotted vertical lineindicates a change in the x-scale. (b) a snapshot from the end ofa single realisation where dispersal distance is 1 and the meanevolved age at death is 31, showing the level of variation andclustering of evolved strategies.

Fig. 3. The evolved dispersal distance after 10 000 generationsfor a range of fixed values age at death. Dispersal mortality is 10throughout. Mean and 95% confidence interval of 10 realisationsof the model are shown.

3.0

Mea

n ev

olve

d di

stan

ce

2.8

2.6

2.4

2.2

2.010 20 30 40 50 60

Age at death70 80 90 100

10 20 30 40 50 60 70 80 90 100

Age at death

18000

19000

20000

21000

22000

23000

Pop

ulat

ion

size

Fig. 4. The mean and 95% confidence interval of populationsize is shown for a range of fixed age at death. Dispersalmortality is 5, dispersal fixed at 1.5, age at death fixed for eachrealisation but varied from 10 to 100 in steps of two. The dottedvertical line indicates the age at death that evolves whendispersal is fixed at 1.5 and cost is 5.

534 OIKOS 113:3 (2006)

reduces the strength of kin selection resulting in selection

for greater potential longevity (Travis 2004).

It may be that many seeds fall very close to the parent

and are unable to germinate (Bullock et al. 2002a). We

have also generated results for when an offspring is not

forced to disperse away from the site occupied by its

parent. With no forced move, mortality is increased for

short-range dispersers and therefore there is an increase

in the evolved dispersal distance, and a corresponding

increase in age of death.

Discussion

Limited dispersal creates intense competition amongst

kin and, when fertility declines with age, this leads to

selection of an adaptive age at death. The model results

described in this paper support those of Travis (2004)

and, in addition, illustrate clearly that the evolution of

dispersal can drive the evolution of programmed age at

death. There are two important differences in the model

used in this paper. Firstly, we use a continuous time

model, whereas in Travis (2004) the model ran in discrete

time. Secondly, we incorporate more realistic dispersal

kernels than used in Travis (2004) while retaining the

simplistic mortality kernel. Despite these differences, the

model results are qualitatively identical to those of Travis

(2004) in predicting a much earlier age of death for

shorter distance dispersal (Fig. 2). Allowing the evolu-

tion of both dispersal and age of death was a natural

extension. The results suggest that, in general, dispersal

will respond faster to a change in selection pressure, and

that longevity will evolve in response to any change in

dispersal.

Our results indicate that dispersal distance can also

evolve in response to age of death. Greater dispersal

distances evolve when longevity is increased (Fig. 3).

While this result is perhaps less striking than the

evolution of ageing in response to dispersal, it neatly

illustrates that the evolution of each of the life history

characteristics is dependent on the other. Why does

increased dispersal distance evolve when longevity is

greater? We suggest that in this model the mechanism is

kin selection or inclusive fitness. If individuals live

longer, competition for space with kin will increase,

and an increase in kin competition will generally lead to

increased dispersal (Ronce et al. 2000, Perrin and

0

0.5

1

1.5

2

2.5

3

0 20 40 60 80 100

Mean evolved age at death

Mea

n ev

olve

d di

sper

sal d

ista

nce

1.55

1.6

1.65

1.7

30 31 32 33 34 35 36

Mean evolved age at death

Mea

n ev

olve

d di

sper

sal d

ista

nce

(a)

(b)

25 30(c)

(d)

35 40 45Age at death

1000

2000

3000

4000

5000

Freq

uenc

y

1.0 1.5 2.0 2.5Dispersal distance

0

2500

5000

7500

Freq

uenc

y

Fig. 5. Evolving both dispersal distance and age at death.Dispersal mortality is set at 5 throughout. Eleven trajectories ofmean evolved values are shown, one symbol every 100 genera-tions for 150 000 generations. Starting points are shown as opencircles. (b) shows that all move to the same equilibrium asshown clearly at this more detailed scale. For a single realisationof the model after 100 000 generations there were over 28 000individuals. (c) shows the frequency distribution of age at deathand (d) dispersal distance with dotted lines indicating the meanvalues.

OIKOS 113:3 (2006) 535

Goudet 2001). Our model has a dispersal rate indepen-

dent of age, although Ronce et al. (1998) showed that

fewer offspring should disperse from an older individual

than a younger one.

Any change in the cost of dispersal, is likely to

ultimately impact both on the dispersal characteristics

and the programmed age of death found in a population.

From Fig. 5 it is clear that the dispersal distance evolves

more rapidly than the age at death. Dispersal evolves to

an equilibrium level and then age at death responds. We

predict that, in nature, any change in cost to dispersal,

for example through landscape alteration, will first affect

dispersal rates and distances first and impact on age at

death much more slowly. As we have previously demon-

strated (Travis and Dytham 1998, 1999) unintelligent

dispersers, such as a plants with wind-dispersed seed, are

likely to evolve shorter dispersal distances in response to

habitat fragmentation, and results from this paper lead

us to predict that habitat loss will eventually lead to

younger age at death for individuals in these popula-

tions. We could also speculate on the effect of habitat

persistence and habitat quality on age at death from

what we know about its effect on dispersal distance

(Travis et al. 1999, Travis 2001, Murrell et al. 2002,

Poethke and Hovestadt 2002, Parvinen 2004). However,

changes in habitat persistence can substantially alter the

nature of the spatial population dynamics and this in

turn can modify the strength of kin selection. We suspect

that the interplay between dispersal, age at death and the

spatial population dynamics may in this instance be

particularly complex, and highlight it as an area where

further theory would be beneficial.

In this paper we have considered the evolution of

dispersal and programmed age of death. Theoretical and

empirical work has together contributed to a major

advance in our understanding of both the causes and

consequences of different dispersal behaviours (Clobert

et al. 2001, Bullock et al. 2002b, Bowler and Benton

2005). Most often, work investigating the evolution of

dispersal, considers the process in the absence of

evolution acting on other life history characteristics.

An important feature of this paper is that it investigates

the evolution of dispersal with another character. In

reality, dispersal is likely to evolve with a wide range of

traits such as competitive ability, mating strategy, age of

maturity and resistance to or virulence of pathogens. For

example dispersal may evolve with kin-recognition

mechanisms (Perrin and Lehmann 2001, Lehmann and

Perrin 2003). Theoretical and empirical work dedicated

to investigating the evolution of different life history

characteristics would be valuable and Mittledorf (2004)

reviews some of the existing studies.

Whereas, in the dispersal literature there is a general

recognition that the evolved strategy depends upon a

suite of drivers acting together, in ageing research the

various theories are considered alternates, and in general

it is not considered that the different processes may act

together to shape the life history of an organism. Our

view is that ageing is likely to be determined by a

combination of these processes, and that the relative

importance of each driver will vary depending upon the

organism concerned and the environment it inhabits. For

example, although adaptive age at death may be

important when only a few individuals reach old age, it

is likely to become more important when survival is

higher. It is also worth considering that if selection

favours a finite lifespan, pleiotropic genes may confer a

selective advantage both because they confer increased

fitness early in life and because they promote ageing

later. Rather than the late-life ‘detrimental effect’ being

viewed as an unfortunate indirect effect it may actually

confer an evolutionary benefit of its own. It is possible

that a gene that has the ‘negative’ late life effect may

be selected over a gene with the same early-life

consequences without the pleiotropic effect. Future

work should seek to integrate traditional and adaptive

theories on ageing.

Another outcome of our model that appears robust is

that the evolved age at death is not necessarily the one

that generates the highest population density (Fig. 4).

Evolution operates at a level below that of the popula-

tion (in this model it acts at both individual and kin

group levels) so there is no reason to expect that the

evolved and optimal (in terms of maximising population

size) strategies should coincide. It is worth comparing

the result shown here with those presented by Parvinen

(2004; see also Olivieri and Gouyon 1997). Parvinen

(2004) clearly illustrates that an evolved dispersal rate

can be either higher or lower than the one that optimises

population size, and that the direction of deviation

depends upon the spatial scale of dispersal.

This paper adds to the recent ideas relating to

adaptive ageing and programmed death (Skulachev

2001, 2002, Mitteldorf 2004, Travis 2004). To date, this

work has consisted of verbal arguments illustrated by

judicious use of empirical evidence (Mitteldorf 2004)

and individual-based simulation models (Travis 2004,

Mitteldorf 2005). Theoretically, programmed death can

be adaptive. The key question now, is to what extent is

the ageing of organisms determined by selection acting

directly on it, and to what extent is it a side effect of

selection acting to maximise fitness early in life. Teasing

apart these two is likely to be challenging, and will

demand a mix of theoretical and experimental work.

Studies on mammals (Broussard et al. 2003, 2005) have

made some progress in determining the relative invest-

ment in reproduction or soma (i.e. longevity) through

life, but sample sizes are inevitably small. We believe that

carefully planned microcosm experiments offer consider-

able potential: organisms such as Caenorhabditis

elegans (Gardner et al. 2004), Drosophila melanogaster

(Kennington et al. 2003) and mites (Benton et al. 2004)

536 OIKOS 113:3 (2006)

are all amenable to microcosm work. They are readily

cultured, and have relatively short generation times

allowing evolutionary processes to be studied in the

confines of the laboratory. In addition it is easy to

acquire genotyes of C. elegans that have different

dispersal and ageing behaviours. Experiments could be

designed to test specific predictions of the models. For

example, to test the idea that population viscosity

determines the evolution of ageing, a simple experiment

could be established involving competition between

different ageing genotypes. The spatial structure of a

set of populations could be manipulated by manually

moving individuals and thus artificially increasing dis-

persal, while in the control the same proportion of

individuals could be picked up and replaced in the same

position. An added complication to the interpretation of

results from such a study is that fertility profile will also

evolve under such selection (Rose 1984).

Ageing research, or gerontology, is booming. While

this has no doubt been fuelled by a desire to better

understand and deal with the consequences of an ageing

human population, considerable progress has been made

that can be applied to a broad range of questions,

including ecological ones. Additionally, ecological in-

sights may help elucidate the importance of various

potential drivers of ageing under different environmental

conditions. We can predict, for example, that habitat

fragmentation will have an impact on evolved age at

death through its impact on dispersal because highly

fragmented populations are likely to be inbred.

Although traditionally quite separate fields there

may be much that gerontologists can learn from

ecologists and vice versa. It is our hope that this paper

will help to stimulate interest in the field of ecological

gerontology.

References

Bartke, A. 2001. Mutations prolong life in flies; implications foraging in mammals. � Trends Endocrinol. Metab. 12: 233�234.

Bengtsson, B. O. 1978. Avoiding inbreeding: at what cost? � J.Theor. Biol. 73: 439�444.

Benton, T. G., Cameron, T. C. and Grant, A. 2004. Populationresponses to perturmbations: predictions and responsesfrom laboratory mite populations. � J. Anim. Ecol. 73:983�995.

Bonsall, M. B. and Mangel, M. 2004. Life-history tradeoffs andecological dynamics in the evolution of longevity. � Proc. R.Soc. Lond. 271: 1143�1150.

Bowler, D. E. and Benton, T. G. 2005. Causes and consequencesof animal dispersal strategies: relating individual behaviourto spatial dynamics. � Biol. Rev. 80: 205�225.

Broussard, D. R., Risch, T. S., Dobson, F. S. et al. 2003.Senescence and age-related reproduction of female Colum-bian ground squirrels. � J. Anim. Ecol. 72: 212�219.

Broussard, D. R., Michener, G. S., Risch, T. S. et al. 2005.Somatic senescence: evidence from female Richardson’sground squirrels. � Oikos 108: 591�601.

Bullock, J., Kenward, R. and Hails, R. 2002a. Dispersalecology. � Blackwell Science.

Bullock, J. M., Moy, I. L., Pywell, R. F. et al. 2002b. Plantdispersal and colonisation processes at local and landscapescales. � In: Bullock, J. M., Kenward, R. E. and Hails, R. S.(eds), Dispersal. Blackwell Science.

Clobert, J., Danchin, E., Dhont, A. A. et al. 2001. Dispersal.� Oxford Univ. Press.

Cockburn, A., Osmond, H. L., Mulder, R. A. et al. 2003.Divorce, dispersal and incest avoidance in the cooperativelybreeding superb fairy-wren Malurus cyaneus. � J. Anim.Ecol. 72: 189�202.

Friedman, D. B. and Johnson, T. E. 1988. Three mutants thatextend both the mean and maximum life span of thenematode, Caenorhabditis elegans, define the age-1 gene.� J. Gerentol. 43: 102�109.

Gardner, M., Gems, D. and Viney, M. 2004. Aging in a veryshort-lived nematode. � Exp. Gerontol. 39: 1267�1276.

Gavrilov, L. A. and Gavrilova, N. S. 2002. Evolutionarytheories of aging and longevity. � Sci. World J. 2: 339�356.

Goldsmith, T. C. 2003. The evolution of aging: how Darwin’sdilemma is affecting your chance for a longer and healthierlife. � Universe Inc, Lincoln, USA.

Guarente, L. and Kenyon, C. 2000. Genetic pathways thatregulate ageing in model organisms. � Nature 408: 255�262.

Hamilton, W. D. and May, R. M. 1977. Dispersal in stablehabitats. � Nature 269: 578�581.

Hendry, A. P., Morbey, Y. E., Berg, O. K. et al. 2004. Adaptivevariation in senescence: reproductive lifespan in a wildsalmon population. � Proc. R. Soc. 271: 259�266.

Holzenberger, M., Dupont, J., Ducos, B. et al. 2003. IGF-1receptor regulates lifespan and resistance to oxidative stressin mice. � Nature 421: 182�187.

Kennington, W. J., Killeen, J. R., Goldstein, D. B. et al. 2003.Rapid laboratory evolution of adult body size in Drosophilamelanogaster in response to humidity and temperature.� Evolution 57: 932�936.

Kenyon, C., Chang, J., Gensch, E. et al. 1993. A C. elegansmutant that lives twice as long as wild type. � Nature 366:461�464.

Kimura, K. D., Tissenbaum, H. A., Liu, Y. et al. 1997. Daf-2,an insulin receptor-like gene that regulates longevity anddiapause in Caenorhabditis elegans. � Science 277: 942�946.

Kirkwood, T. B. L. 1977. Evolution of ageing. � Nature 270:301�304.

Kirkwood, T. B. L. 2005. Understanding the odd science ofaging. � Cell 120: 437�447.

Lehmann, L. and Perrin, N. 2003. Inbreeding avoidancethrough kin recognition. Choosy females boost maledispersal. � Am. Nat. 162: 638�652.

Lithgow, G. J. and Gill, M. S. 2003. Cost-free longevity in mice?� Nature 421: 125�126.

Medawar, P. B. 1952. An unsolved problem of biology. � H. K.Lewis.

Migliaccio, E., Giorgio, M., Mele, S. et al. 1999. The p66shc

adaptor protein controls oxidative stress response andlifespan in mammals. � Nature 402: 309�313.

Mitteldorf, J. 2004. Ageing selected for its own sake. � Evol.Ecol. Res. 6: 937�953.

Mitteldorf, J. 2006. Demographic homeostasis and the evolu-tion of senescence. � Evol. Ecol. Res. (in press).

Murrell, D. J., Travis, J. M. J. and Dytham, C. 2002. Theevolution of dispersal distance in spatially-structured popu-lations. � Oikos 97: 229�236.

Olivieri, I. and Gouyon, P. H. 1997. Evolution of migration rateand other traits: the metapopulation effect. � In: Hanski, I.and Gilpin, M.E. (eds), Metapopulation biology: ecology,genetics and evolution. Academic Press, pp. 293�323.

Olshansky, S. J., Halflick, L. and Carnes, B. A. 2002. The truthabout human aging [position statement on human aging].� J. Gerontol. 57A: B292�B297.

OIKOS 113:3 (2006) 537

Pearl, R. 1922. The biology of death. � J. B. Lippincott,Philadelphia.

Palmer, J. B. 1992. Hierarchical and concurrent individual basedmodelling. � In: DeAngelis, D. L. and Gross, L. J. (eds),Individual-based models and approaches in ecology:populations, communities and ecosystems. Chapman andHall, pp. 188�207.

Parvinen, K. 2004. Adaptive responses to landscape distur-bances: theory. � In: Ferriere, R., Dieckmann, U. andCouvet, D. (eds), Evolutionary conservation biology.� Cambridge Univ. Press, pp. 265�286.

Perrin, N. and Mazalov, V. 1999. Dispersal and inbreedingavoidance. � Am. Nat. 154: 282�292.

Perrin, N. and Goudet, J. 2001. Inbreeding, kinship and theevolution of natal dispersal. � In: Clobert, J. et al. (eds),Dispersal. Oxford Univ. Press, pp. 155�167.

Perrin, N. and Lehmann, L. 2001. Is sociality driven by thecosts of dispersal or the benefits of philopatry? A role forkin-discrimination mechanisms. � Am. Nat. 158: 471�583.

Poethke, H. J. and Hovestadt, T. 2002. Evolution of density andpatch-size-dependent dispersal rates. � Proc. R. Soc. Lond.269: 637�645.

Renshaw, E. 1991. Modelling biological populations in spaceand time. � Cambridge Univ. Press.

Roff, D. A. 1986. Evolution of wing polymorphism and itsimpact on life cycle adaptation in insects. � In: Taylor, F.and Karban, K. (eds), The evolution of insect life cycles.Springer-Verlag, pp. 204�221.

Roff, D. A. 1994. The evolution of flightlessness: is historyimportant? � Evol. Ecol. 8: 629�657.

Ronce, O., Clobert, J. and Massot, M. 1998. Natal dispersal andsenescence. � Proc. Nat. Acad. Sci. 95: 600�605.

Ronce, O., Gandon, S. and Rousset, F. 2000. Kin selection andnatal dispersal in an age-structured population. � Theor.Popul. Biol. 58: 143�159.

Rose, M. R. 1984. Laboratory evolution of postponed senes-cence in Drosophila melanogaster. � Evolution 38: 1004�1010.

Skulachev, V. P. 2001. The programmed death phenomena,aging, and the Samurai law of biology. � Exp. Gerontol. 36:995�1024.

Skulachev, V. P. 2002. Programmed death phenomena: fromorganelle to organism. � Ann. N. Y. Acad. Sci. 959: 214�237.

Tatar, M., Kopelman, A., Epstein, D. et al. 2001. A mutantDrosophila insulin receptor homolog that extends life-spanand impairs neuroendocrine function. � Science 292: 107�110.

Travis, J. M. J. 2001. The color of noise and the evolution ofdispersal. � Ecol. Res. 16: 157�163.

Travis, J. M. J. 2004. The evolution of programmed death in aspatially structured population. � J. Gerontol. 59A: 301�305.

Travis, J. M. J. and Dytham, C. 1998. The evolution of dispersalin a metapopulation: a spatially explicit, individual-basedmodel. � Proc. R. Soc. Lond. 265: 17�23.

Travis, J. M. J. and Dytham, C. 1999. Habitat persistence,habitat availability and the evolution of dispersal. � Proc. R.Soc. Lond. 266: 723�728.

Travis, J. M. J., Murrell, D. J. and Dytham, C. 1999. Theevolution of density-dependent dispersal. � Proc. R. Soc.Lond. 266: 1837�1842.

Wallace, A. R. 1889. The action of natural selection inproducing old age, decay and death. A note by Wallacewritten ‘some time between 1865 and 1870’ and included inWeismann, A., Essays upon heredity and kindred biologicalproblems. � Clarendon Press.

Weismann, A. 1882. Uber die Dauer des Lebens. � Verlag vonGustav Fisher, Jena.

Williams, G. C. 1957. Pleiotropy, natural selection and theevolution of senescence. � Evolution 11: 398�411.

538 OIKOS 113:3 (2006)